Yes: K2-Horizon-3.7B runs on all 54 Macs in the LLMCheck index. It needs about 4 GB of unified memory at a 4K context and about 14 GB at 64K. The smallest and cheapest current Mac that runs it at 15+ tok/s is the Mac mini M6 16 GB ($899; ~39 tok/s est.). No measured run yet.
One row per Mac configuration, grouped by product line. The build is the one each speed is for; the context column is the longest prompt that fits inside macOS’s default GPU memory limit with a 16-bit KV cache (“raised” where the limit has to be raised first; “full” when the whole window fits). Speeds are decode tok/s from the LLMCheck index, each with its provenance mark. For dense models, the formula’s typical error is about 9% (the median gap to 75 measured plain-decode runs; it tends to undershoot (median −6%), and an estimate scaled from the model’s own measured run is usually closer; method).
| Mac | Verdict | Build | Speed | Max context | Model page |
|---|---|---|---|---|---|
| MacBook Air | |||||
| MacBook Air M5 16 GB | Runs | Q4_K_M | estimated39 tok/s | 55K | — |
| MacBook Air M5 24 GB | Runs | Q4_K_M | estimated39 tok/s | 95K | — |
| MacBook Air M5 32 GB | Runs | Q4_K_M | estimated39 tok/s | 142K | — |
| MacBook Air M1 8 GB refurb | Runs | Q4_K_M | estimated18 tok/s | 12K | — |
| MacBook Air M4 16 GB used | Runs | Q4_K_M | estimated31 tok/s | 48K | — |
| MacBook Air M2 16 GB refurb | Runs | Q4_K_M | estimated26 tok/s | 48K | — |
| MacBook Air M3 16 GB refurb | Runs | Q4_K_M | estimated26 tok/s | 48K | — |
| MacBook Air M1 16 GB refurb | Runs | Q4_K_M | estimated18 tok/s | 48K | — |
| MacBook Air M4 24 GB used | Runs | Q4_K_M | estimated31 tok/s | 83K | — |
| MacBook Air M2 24 GB refurb | Runs | Q4_K_M | estimated26 tok/s | 83K | — |
| MacBook Air M3 24 GB refurb | Runs | Q4_K_M | estimated26 tok/s | 83K | — |
| MacBook Pro | |||||
| MacBook Pro M6 16 GB | Runs | Q4_K_M | estimated39 tok/s | 55K | — |
| MacBook Pro M5 Pro 24 GB | Runs | Q4_K_M | estimated74 tok/s | 95K | — |
| MacBook Pro M6 24 GB | Runs | Q4_K_M | estimated43 tok/s | 95K | — |
| MacBook Pro M6 32 GB | Runs | Q4_K_M | estimated43 tok/s | 142K | — |
| MacBook Pro M5 Pro 48 GB | Runs | Q4_K_M | estimated74 tok/s | 225K | — |
| MacBook Pro M5 Max 64 GB | Runs | Q4_K_M | estimated133 tok/s | 320K | M5 Max page |
| MacBook Pro M5 Pro 64 GB | Runs | Q4_K_M | estimated74 tok/s | 320K | — |
| MacBook Pro M5 Max 128 GB | Runs | Q4_K_M | estimated133 tok/s | 512K (full) | M5 Max page |
| MacBook Pro M3 Max 64 GB refurb | Runs | Q4_K_M | estimated93 tok/s | 295K | — |
| MacBook Pro M3 Max 128 GB refurb | Runs | Q4_K_M | estimated93 tok/s | 512K (full) | — |
| Mac mini | |||||
| Mac mini M6 16 GB | Runs | Q4_K_M | estimated39 tok/s | 55K | — |
| Mac mini M5 Pro 24 GB | Runs | Q4_K_M | estimated74 tok/s | 95K | — |
| Mac mini M6 24 GB | Runs | Q4_K_M | estimated43 tok/s | 95K | — |
| Mac mini M6 32 GB | Runs | Q4_K_M | estimated43 tok/s | 142K | — |
| Mac mini M5 Pro 48 GB | Runs | Q4_K_M | estimated74 tok/s | 225K | — |
| Mac mini M5 Pro 64 GB | Runs | Q4_K_M | estimated74 tok/s | 320K | — |
| Mac mini M2 Pro 16 GB refurb | Runs | Q4_K_M | estimated50 tok/s | 48K | — |
| Mac mini M4 16 GB used | Runs | Q4_K_M | estimated31 tok/s | 48K | — |
| Mac mini M4 Pro 24 GB used | Runs | Q4_K_M | estimated66 tok/s | 83K | — |
| Mac mini M4 24 GB used | Runs | Q4_K_M | estimated31 tok/s | 83K | — |
| Mac mini M2 Pro 32 GB refurb | Runs | Q4_K_M | estimated50 tok/s | 118K | — |
| Mac mini M4 Pro 48 GB used | Runs | Q4_K_M | estimated66 tok/s | 215K | — |
| Mac mini M4 Pro 64 GB used | Runs | Q4_K_M | estimated66 tok/s | 295K | — |
| iMac | |||||
| iMac M6 16 GB | Runs | Q4_K_M | estimated39 tok/s | 55K | — |
| iMac M6 24 GB | Runs | Q4_K_M | estimated43 tok/s | 95K | — |
| iMac M6 32 GB | Runs | Q4_K_M | estimated43 tok/s | 142K | — |
| Mac Studio | |||||
| Mac Studio M5 Max 36 GB | Runs | Q4_K_M | estimated105 tok/s | 163K | M5 Max page |
| Mac Studio M5 Max 48 GB | Runs | Q4_K_M | estimated133 tok/s | 225K | M5 Max page |
| Mac Studio M5 Max 64 GB | Runs | Q4_K_M | estimated133 tok/s | 320K | M5 Max page |
| Mac Studio M5 Ultra 96 GB | Runs | Q4_K_M | estimated223 tok/s | 492K | — |
| Mac Studio M5 Max 128 GB | Runs | Q4_K_M | estimated133 tok/s | 512K (full) | M5 Max page |
| Mac Studio M5 Ultra 256 GB | Runs | Q4_K_M | estimated223 tok/s | 512K (full) | — |
| Mac Studio M5 Ultra 512 GB | Runs | Q4_K_M | estimated223 tok/s | 512K (full) | — |
| Mac Studio M2 Max 32 GB refurb | Runs | Q4_K_M | estimated93 tok/s | 118K | — |
| Mac Studio M1 Ultra 64 GB refurb | Runs | Q4_K_M | estimated164 tok/s | 295K | — |
| Mac Studio M4 Max 64 GB used | Runs | Q4_K_M | estimated121 tok/s | 295K | — |
| Mac Studio M2 Max 64 GB refurb | Runs | Q4_K_M | estimated93 tok/s | 295K | — |
| Mac Studio M1 Ultra 128 GB refurb | Runs | Q4_K_M | estimated164 tok/s | 512K (full) | — |
| Mac Studio M2 Ultra 128 GB refurb | Runs | Q4_K_M | estimated164 tok/s | 512K (full) | — |
| Mac Studio M4 Max 128 GB used | Runs | Q4_K_M | estimated121 tok/s | 512K (full) | — |
| Mac Studio M2 Ultra 192 GB refurb | Runs | Q4_K_M | estimated164 tok/s | 512K (full) | — |
| Mac Studio M3 Ultra 256 GB used | Runs | Q4_K_M | estimated167 tok/s | 512K (full) | — |
| Mac Studio M3 Ultra 512 GB refurb | Runs | Q4_K_M | estimated167 tok/s | 512K (full) | — |
Context memory is computed from its config.json (GQA on all 36 layers (8 kv heads x 128)): about 9.7 GB of 16-bit KV cache at 64K tokens. A q8 cache (llama.cpp -ctk q8_0 -ctv q8_0, Ollama OLLAMA_KV_CACHE_TYPE=q8_0) halves that part. Runtimes also grow a compute buffer with context, so a fit within about 1 GB of the limit is optimistic.
macOS lets the GPU use only part of unified memory by default. K2-Horizon-3.7B fits inside that default limit on every Mac it fits at all, so no Mac in the table needs it raised at a short context. Long contexts may still need it: the context column already uses each Mac’s default limit.
Weights are 5.1B parameters at each format’s effective size (the index’s GB-per-billion table), plus about 2 GB for the runtime and a short context; longer contexts add the KV cache computed from the model’s config.json at 16 bits. The first row is the catalog’s own figure (about 4 GB at a 4K context), the one every verdict on this page uses. “Smallest Mac” is the smallest memory size among the Macs Apple sells new that holds it at a 4K context inside that Mac’s default GPU limit. Estimates, not measurements.
| Build | Weights | RAM at 4K | at 32K | at 64K | Smallest Mac |
|---|---|---|---|---|---|
| As catalogued | — | 4 GB | 8 GB | 13 GB | 16 GB |
| Q2_K | 1.7 GB | 4 GB | 8 GB | 13 GB | 16 GB |
| Q3_K_M | 2.5 GB | 4 GB | 9 GB | 14 GB | 16 GB |
| Q4_K_M | 2.9 GB | 5 GB | 9 GB | 14 GB | 16 GB |
| Q5_K_M | 3.6 GB | 6 GB | 10 GB | 15 GB | 16 GB |
| Q6_K | 4.2 GB | 6 GB | 10 GB | 15 GB | 16 GB |
| Q8_0 | 5.4 GB | 7 GB | 12 GB | 16 GB | 16 GB |
| F16 | 10.2 GB | 12 GB | 16 GB | 21 GB | 24 GB |
The index holds no published or community run of K2-Horizon-3.7B yet, so every speed on this page is an estimate from the memory-bandwidth model. LLMCheck runs no tests of its own: if you have run it on a Mac, submit the figure with its setup →
The index has no verified Ollama tag for K2-Horizon-3.7B. Start from its listing and copy the exact name of a build that fits: Hugging Face repo → · model card →
GGUF builds run in LM Studio or llama.cpp; MLX builds run in LM Studio or mlx-lm.
On the Mac mini M6 (16 GB), the smallest current Mac that runs it inside the default GPU limit, the index figure is ~39 tok/s (estimated) at Q4_K_M; keep the context at or under 55K there, or use a q8 KV cache.
Step-by-step: the Ollama install guide.
Mac mini M6 (16 GB) — smallest current Mac that runs it and cheapest at 15+ tok/s, $899 new (Apple list price, checked 2026-11-14), ~39 tok/s (estimated).
As an Amazon Associate, LLMCheck earns from qualifying purchases. Affiliate links cost you nothing extra and never influence the verdicts.
Yes. K2-Horizon-3.7B needs about 4 GB of unified memory at a short (4K-token) context, so it runs on 54 of the 54 Mac configurations in the LLMCheck index. The smallest Mac Apple sells new that runs it is the Mac mini M6 (16 GB).
About 4 GB at a 4K context: the weights plus about 2 GB for the runtime. Context adds KV cache: about 9 GB in total at 32K and 14 GB at 64K with the default 16-bit cache (about 9 GB at 64K with a q8 cache). These are estimates from the model’s config.json, not measurements.
On an M5 Max, the chip the leaderboard quotes it on, the index figure is ~133 tok/s (estimated) at Q4_K_M. No published or community run is in the index yet; every speed is an estimate from the memory-bandwidth model. Every Mac’s figure is in the table on this page with its provenance: published, community run or estimated.
Among Macs Apple sells new, the Mac mini M6 (16 GB), $899 new (Apple list price, checked 2026-11-14), is the cheapest that reaches 15 tok/s with K2-Horizon-3.7B, at ~39 tok/s (estimated).
No. On all 54 Mac configurations where K2-Horizon-3.7B fits, it fits inside that Mac’s default GPU memory limit at a short context; only a long context on the smallest of them can push it over.
Q4_K_M (about 3 GB of weights) is the common 4-bit GGUF choice. Q8_0 (about 5 GB) is closer to full quality and needs a 16 GB Mac at the default GPU limit. Q3_K_M and Q2_K shrink it to about 2 and 2 GB at a quality cost. The index’s reference figure (133 tok/s on an M5 Max) is for its Q4_K_M build.
Its own window is 512K tokens. At macOS’s default GPU limit, the Mac mini M6 (16 GB) holds about 55K. The full window needs about 81 GB, which a 128 GB Mac holds at the default limit. A q8 KV cache roughly doubles these lengths.
LLMCheck. (2026). Can I run K2-Horizon-3.7B on a Mac? llmcheck.net. https://llmcheck.net/can-i-run/k2-horizon-37b/ (updated ). CC BY 4.0.